1 #ifndef _LINUX_MM_H 2 #define _LINUX_MM_H 3 4 #include <linux/errno.h> 5 6 #ifdef __KERNEL__ 7 8 #include <linux/gfp.h> 9 #include <linux/list.h> 10 #include <linux/mmzone.h> 11 #include <linux/rbtree.h> 12 #include <linux/prio_tree.h> 13 #include <linux/debug_locks.h> 14 #include <linux/mm_types.h> 15 #include <linux/range.h> 16 17 struct mempolicy; 18 struct anon_vma; 19 struct file_ra_state; 20 struct user_struct; 21 struct writeback_control; 22 struct rlimit; 23 24 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */ 25 extern unsigned long max_mapnr; 26 #endif 27 28 extern unsigned long num_physpages; 29 extern unsigned long totalram_pages; 30 extern void * high_memory; 31 extern int page_cluster; 32 33 #ifdef CONFIG_SYSCTL 34 extern int sysctl_legacy_va_layout; 35 #else 36 #define sysctl_legacy_va_layout 0 37 #endif 38 39 #include <asm/page.h> 40 #include <asm/pgtable.h> 41 #include <asm/processor.h> 42 43 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) 44 45 /* to align the pointer to the (next) page boundary */ 46 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) 47 48 /* 49 * Linux kernel virtual memory manager primitives. 50 * The idea being to have a "virtual" mm in the same way 51 * we have a virtual fs - giving a cleaner interface to the 52 * mm details, and allowing different kinds of memory mappings 53 * (from shared memory to executable loading to arbitrary 54 * mmap() functions). 55 */ 56 57 extern struct kmem_cache *vm_area_cachep; 58 59 #ifndef CONFIG_MMU 60 extern struct rb_root nommu_region_tree; 61 extern struct rw_semaphore nommu_region_sem; 62 63 extern unsigned int kobjsize(const void *objp); 64 #endif 65 66 /* 67 * vm_flags in vm_area_struct, see mm_types.h. 68 */ 69 #define VM_READ 0x00000001 /* currently active flags */ 70 #define VM_WRITE 0x00000002 71 #define VM_EXEC 0x00000004 72 #define VM_SHARED 0x00000008 73 74 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ 75 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ 76 #define VM_MAYWRITE 0x00000020 77 #define VM_MAYEXEC 0x00000040 78 #define VM_MAYSHARE 0x00000080 79 80 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ 81 #define VM_GROWSUP 0x00000200 82 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ 83 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ 84 85 #define VM_EXECUTABLE 0x00001000 86 #define VM_LOCKED 0x00002000 87 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ 88 89 /* Used by sys_madvise() */ 90 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ 91 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ 92 93 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ 94 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ 95 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */ 96 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ 97 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ 98 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ 99 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */ 100 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */ 101 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */ 102 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */ 103 104 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */ 105 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ 106 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */ 107 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */ 108 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ 109 110 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ 111 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS 112 #endif 113 114 #ifdef CONFIG_STACK_GROWSUP 115 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 116 #else 117 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) 118 #endif 119 120 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ) 121 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK 122 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK)) 123 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ) 124 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ) 125 126 /* 127 * special vmas that are non-mergable, non-mlock()able 128 */ 129 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP) 130 131 /* 132 * mapping from the currently active vm_flags protection bits (the 133 * low four bits) to a page protection mask.. 134 */ 135 extern pgprot_t protection_map[16]; 136 137 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ 138 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */ 139 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */ 140 141 /* 142 * This interface is used by x86 PAT code to identify a pfn mapping that is 143 * linear over entire vma. This is to optimize PAT code that deals with 144 * marking the physical region with a particular prot. This is not for generic 145 * mm use. Note also that this check will not work if the pfn mapping is 146 * linear for a vma starting at physical address 0. In which case PAT code 147 * falls back to slow path of reserving physical range page by page. 148 */ 149 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma) 150 { 151 return (vma->vm_flags & VM_PFN_AT_MMAP); 152 } 153 154 static inline int is_pfn_mapping(struct vm_area_struct *vma) 155 { 156 return (vma->vm_flags & VM_PFNMAP); 157 } 158 159 /* 160 * vm_fault is filled by the the pagefault handler and passed to the vma's 161 * ->fault function. The vma's ->fault is responsible for returning a bitmask 162 * of VM_FAULT_xxx flags that give details about how the fault was handled. 163 * 164 * pgoff should be used in favour of virtual_address, if possible. If pgoff 165 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear 166 * mapping support. 167 */ 168 struct vm_fault { 169 unsigned int flags; /* FAULT_FLAG_xxx flags */ 170 pgoff_t pgoff; /* Logical page offset based on vma */ 171 void __user *virtual_address; /* Faulting virtual address */ 172 173 struct page *page; /* ->fault handlers should return a 174 * page here, unless VM_FAULT_NOPAGE 175 * is set (which is also implied by 176 * VM_FAULT_ERROR). 177 */ 178 }; 179 180 /* 181 * These are the virtual MM functions - opening of an area, closing and 182 * unmapping it (needed to keep files on disk up-to-date etc), pointer 183 * to the functions called when a no-page or a wp-page exception occurs. 184 */ 185 struct vm_operations_struct { 186 void (*open)(struct vm_area_struct * area); 187 void (*close)(struct vm_area_struct * area); 188 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); 189 190 /* notification that a previously read-only page is about to become 191 * writable, if an error is returned it will cause a SIGBUS */ 192 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); 193 194 /* called by access_process_vm when get_user_pages() fails, typically 195 * for use by special VMAs that can switch between memory and hardware 196 */ 197 int (*access)(struct vm_area_struct *vma, unsigned long addr, 198 void *buf, int len, int write); 199 #ifdef CONFIG_NUMA 200 /* 201 * set_policy() op must add a reference to any non-NULL @new mempolicy 202 * to hold the policy upon return. Caller should pass NULL @new to 203 * remove a policy and fall back to surrounding context--i.e. do not 204 * install a MPOL_DEFAULT policy, nor the task or system default 205 * mempolicy. 206 */ 207 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); 208 209 /* 210 * get_policy() op must add reference [mpol_get()] to any policy at 211 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure 212 * in mm/mempolicy.c will do this automatically. 213 * get_policy() must NOT add a ref if the policy at (vma,addr) is not 214 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. 215 * If no [shared/vma] mempolicy exists at the addr, get_policy() op 216 * must return NULL--i.e., do not "fallback" to task or system default 217 * policy. 218 */ 219 struct mempolicy *(*get_policy)(struct vm_area_struct *vma, 220 unsigned long addr); 221 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from, 222 const nodemask_t *to, unsigned long flags); 223 #endif 224 }; 225 226 struct mmu_gather; 227 struct inode; 228 229 #define page_private(page) ((page)->private) 230 #define set_page_private(page, v) ((page)->private = (v)) 231 232 /* 233 * FIXME: take this include out, include page-flags.h in 234 * files which need it (119 of them) 235 */ 236 #include <linux/page-flags.h> 237 238 /* 239 * Methods to modify the page usage count. 240 * 241 * What counts for a page usage: 242 * - cache mapping (page->mapping) 243 * - private data (page->private) 244 * - page mapped in a task's page tables, each mapping 245 * is counted separately 246 * 247 * Also, many kernel routines increase the page count before a critical 248 * routine so they can be sure the page doesn't go away from under them. 249 */ 250 251 /* 252 * Drop a ref, return true if the refcount fell to zero (the page has no users) 253 */ 254 static inline int put_page_testzero(struct page *page) 255 { 256 VM_BUG_ON(atomic_read(&page->_count) == 0); 257 return atomic_dec_and_test(&page->_count); 258 } 259 260 /* 261 * Try to grab a ref unless the page has a refcount of zero, return false if 262 * that is the case. 263 */ 264 static inline int get_page_unless_zero(struct page *page) 265 { 266 return atomic_inc_not_zero(&page->_count); 267 } 268 269 extern int page_is_ram(unsigned long pfn); 270 271 /* Support for virtually mapped pages */ 272 struct page *vmalloc_to_page(const void *addr); 273 unsigned long vmalloc_to_pfn(const void *addr); 274 275 /* 276 * Determine if an address is within the vmalloc range 277 * 278 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there 279 * is no special casing required. 280 */ 281 static inline int is_vmalloc_addr(const void *x) 282 { 283 #ifdef CONFIG_MMU 284 unsigned long addr = (unsigned long)x; 285 286 return addr >= VMALLOC_START && addr < VMALLOC_END; 287 #else 288 return 0; 289 #endif 290 } 291 #ifdef CONFIG_MMU 292 extern int is_vmalloc_or_module_addr(const void *x); 293 #else 294 static inline int is_vmalloc_or_module_addr(const void *x) 295 { 296 return 0; 297 } 298 #endif 299 300 static inline struct page *compound_head(struct page *page) 301 { 302 if (unlikely(PageTail(page))) 303 return page->first_page; 304 return page; 305 } 306 307 static inline int page_count(struct page *page) 308 { 309 return atomic_read(&compound_head(page)->_count); 310 } 311 312 static inline void get_page(struct page *page) 313 { 314 page = compound_head(page); 315 VM_BUG_ON(atomic_read(&page->_count) == 0); 316 atomic_inc(&page->_count); 317 } 318 319 static inline struct page *virt_to_head_page(const void *x) 320 { 321 struct page *page = virt_to_page(x); 322 return compound_head(page); 323 } 324 325 /* 326 * Setup the page count before being freed into the page allocator for 327 * the first time (boot or memory hotplug) 328 */ 329 static inline void init_page_count(struct page *page) 330 { 331 atomic_set(&page->_count, 1); 332 } 333 334 void put_page(struct page *page); 335 void put_pages_list(struct list_head *pages); 336 337 void split_page(struct page *page, unsigned int order); 338 339 /* 340 * Compound pages have a destructor function. Provide a 341 * prototype for that function and accessor functions. 342 * These are _only_ valid on the head of a PG_compound page. 343 */ 344 typedef void compound_page_dtor(struct page *); 345 346 static inline void set_compound_page_dtor(struct page *page, 347 compound_page_dtor *dtor) 348 { 349 page[1].lru.next = (void *)dtor; 350 } 351 352 static inline compound_page_dtor *get_compound_page_dtor(struct page *page) 353 { 354 return (compound_page_dtor *)page[1].lru.next; 355 } 356 357 static inline int compound_order(struct page *page) 358 { 359 if (!PageHead(page)) 360 return 0; 361 return (unsigned long)page[1].lru.prev; 362 } 363 364 static inline void set_compound_order(struct page *page, unsigned long order) 365 { 366 page[1].lru.prev = (void *)order; 367 } 368 369 /* 370 * Multiple processes may "see" the same page. E.g. for untouched 371 * mappings of /dev/null, all processes see the same page full of 372 * zeroes, and text pages of executables and shared libraries have 373 * only one copy in memory, at most, normally. 374 * 375 * For the non-reserved pages, page_count(page) denotes a reference count. 376 * page_count() == 0 means the page is free. page->lru is then used for 377 * freelist management in the buddy allocator. 378 * page_count() > 0 means the page has been allocated. 379 * 380 * Pages are allocated by the slab allocator in order to provide memory 381 * to kmalloc and kmem_cache_alloc. In this case, the management of the 382 * page, and the fields in 'struct page' are the responsibility of mm/slab.c 383 * unless a particular usage is carefully commented. (the responsibility of 384 * freeing the kmalloc memory is the caller's, of course). 385 * 386 * A page may be used by anyone else who does a __get_free_page(). 387 * In this case, page_count still tracks the references, and should only 388 * be used through the normal accessor functions. The top bits of page->flags 389 * and page->virtual store page management information, but all other fields 390 * are unused and could be used privately, carefully. The management of this 391 * page is the responsibility of the one who allocated it, and those who have 392 * subsequently been given references to it. 393 * 394 * The other pages (we may call them "pagecache pages") are completely 395 * managed by the Linux memory manager: I/O, buffers, swapping etc. 396 * The following discussion applies only to them. 397 * 398 * A pagecache page contains an opaque `private' member, which belongs to the 399 * page's address_space. Usually, this is the address of a circular list of 400 * the page's disk buffers. PG_private must be set to tell the VM to call 401 * into the filesystem to release these pages. 402 * 403 * A page may belong to an inode's memory mapping. In this case, page->mapping 404 * is the pointer to the inode, and page->index is the file offset of the page, 405 * in units of PAGE_CACHE_SIZE. 406 * 407 * If pagecache pages are not associated with an inode, they are said to be 408 * anonymous pages. These may become associated with the swapcache, and in that 409 * case PG_swapcache is set, and page->private is an offset into the swapcache. 410 * 411 * In either case (swapcache or inode backed), the pagecache itself holds one 412 * reference to the page. Setting PG_private should also increment the 413 * refcount. The each user mapping also has a reference to the page. 414 * 415 * The pagecache pages are stored in a per-mapping radix tree, which is 416 * rooted at mapping->page_tree, and indexed by offset. 417 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space 418 * lists, we instead now tag pages as dirty/writeback in the radix tree. 419 * 420 * All pagecache pages may be subject to I/O: 421 * - inode pages may need to be read from disk, 422 * - inode pages which have been modified and are MAP_SHARED may need 423 * to be written back to the inode on disk, 424 * - anonymous pages (including MAP_PRIVATE file mappings) which have been 425 * modified may need to be swapped out to swap space and (later) to be read 426 * back into memory. 427 */ 428 429 /* 430 * The zone field is never updated after free_area_init_core() 431 * sets it, so none of the operations on it need to be atomic. 432 */ 433 434 435 /* 436 * page->flags layout: 437 * 438 * There are three possibilities for how page->flags get 439 * laid out. The first is for the normal case, without 440 * sparsemem. The second is for sparsemem when there is 441 * plenty of space for node and section. The last is when 442 * we have run out of space and have to fall back to an 443 * alternate (slower) way of determining the node. 444 * 445 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS | 446 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS | 447 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS | 448 */ 449 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 450 #define SECTIONS_WIDTH SECTIONS_SHIFT 451 #else 452 #define SECTIONS_WIDTH 0 453 #endif 454 455 #define ZONES_WIDTH ZONES_SHIFT 456 457 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS 458 #define NODES_WIDTH NODES_SHIFT 459 #else 460 #ifdef CONFIG_SPARSEMEM_VMEMMAP 461 #error "Vmemmap: No space for nodes field in page flags" 462 #endif 463 #define NODES_WIDTH 0 464 #endif 465 466 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ 467 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) 468 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) 469 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) 470 471 /* 472 * We are going to use the flags for the page to node mapping if its in 473 * there. This includes the case where there is no node, so it is implicit. 474 */ 475 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0) 476 #define NODE_NOT_IN_PAGE_FLAGS 477 #endif 478 479 #ifndef PFN_SECTION_SHIFT 480 #define PFN_SECTION_SHIFT 0 481 #endif 482 483 /* 484 * Define the bit shifts to access each section. For non-existant 485 * sections we define the shift as 0; that plus a 0 mask ensures 486 * the compiler will optimise away reference to them. 487 */ 488 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) 489 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) 490 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) 491 492 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */ 493 #ifdef NODE_NOT_IN_PAGEFLAGS 494 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) 495 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ 496 SECTIONS_PGOFF : ZONES_PGOFF) 497 #else 498 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) 499 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ 500 NODES_PGOFF : ZONES_PGOFF) 501 #endif 502 503 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) 504 505 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 506 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS 507 #endif 508 509 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) 510 #define NODES_MASK ((1UL << NODES_WIDTH) - 1) 511 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) 512 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) 513 514 static inline enum zone_type page_zonenum(struct page *page) 515 { 516 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; 517 } 518 519 /* 520 * The identification function is only used by the buddy allocator for 521 * determining if two pages could be buddies. We are not really 522 * identifying a zone since we could be using a the section number 523 * id if we have not node id available in page flags. 524 * We guarantee only that it will return the same value for two 525 * combinable pages in a zone. 526 */ 527 static inline int page_zone_id(struct page *page) 528 { 529 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; 530 } 531 532 static inline int zone_to_nid(struct zone *zone) 533 { 534 #ifdef CONFIG_NUMA 535 return zone->node; 536 #else 537 return 0; 538 #endif 539 } 540 541 #ifdef NODE_NOT_IN_PAGE_FLAGS 542 extern int page_to_nid(struct page *page); 543 #else 544 static inline int page_to_nid(struct page *page) 545 { 546 return (page->flags >> NODES_PGSHIFT) & NODES_MASK; 547 } 548 #endif 549 550 static inline struct zone *page_zone(struct page *page) 551 { 552 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; 553 } 554 555 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) 556 static inline unsigned long page_to_section(struct page *page) 557 { 558 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; 559 } 560 #endif 561 562 static inline void set_page_zone(struct page *page, enum zone_type zone) 563 { 564 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); 565 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; 566 } 567 568 static inline void set_page_node(struct page *page, unsigned long node) 569 { 570 page->flags &= ~(NODES_MASK << NODES_PGSHIFT); 571 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; 572 } 573 574 static inline void set_page_section(struct page *page, unsigned long section) 575 { 576 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); 577 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; 578 } 579 580 static inline void set_page_links(struct page *page, enum zone_type zone, 581 unsigned long node, unsigned long pfn) 582 { 583 set_page_zone(page, zone); 584 set_page_node(page, node); 585 set_page_section(page, pfn_to_section_nr(pfn)); 586 } 587 588 /* 589 * Some inline functions in vmstat.h depend on page_zone() 590 */ 591 #include <linux/vmstat.h> 592 593 static __always_inline void *lowmem_page_address(struct page *page) 594 { 595 return __va(page_to_pfn(page) << PAGE_SHIFT); 596 } 597 598 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) 599 #define HASHED_PAGE_VIRTUAL 600 #endif 601 602 #if defined(WANT_PAGE_VIRTUAL) 603 #define page_address(page) ((page)->virtual) 604 #define set_page_address(page, address) \ 605 do { \ 606 (page)->virtual = (address); \ 607 } while(0) 608 #define page_address_init() do { } while(0) 609 #endif 610 611 #if defined(HASHED_PAGE_VIRTUAL) 612 void *page_address(struct page *page); 613 void set_page_address(struct page *page, void *virtual); 614 void page_address_init(void); 615 #endif 616 617 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) 618 #define page_address(page) lowmem_page_address(page) 619 #define set_page_address(page, address) do { } while(0) 620 #define page_address_init() do { } while(0) 621 #endif 622 623 /* 624 * On an anonymous page mapped into a user virtual memory area, 625 * page->mapping points to its anon_vma, not to a struct address_space; 626 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h. 627 * 628 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled, 629 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit; 630 * and then page->mapping points, not to an anon_vma, but to a private 631 * structure which KSM associates with that merged page. See ksm.h. 632 * 633 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used. 634 * 635 * Please note that, confusingly, "page_mapping" refers to the inode 636 * address_space which maps the page from disk; whereas "page_mapped" 637 * refers to user virtual address space into which the page is mapped. 638 */ 639 #define PAGE_MAPPING_ANON 1 640 #define PAGE_MAPPING_KSM 2 641 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM) 642 643 extern struct address_space swapper_space; 644 static inline struct address_space *page_mapping(struct page *page) 645 { 646 struct address_space *mapping = page->mapping; 647 648 VM_BUG_ON(PageSlab(page)); 649 if (unlikely(PageSwapCache(page))) 650 mapping = &swapper_space; 651 else if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON)) 652 mapping = NULL; 653 return mapping; 654 } 655 656 /* Neutral page->mapping pointer to address_space or anon_vma or other */ 657 static inline void *page_rmapping(struct page *page) 658 { 659 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS); 660 } 661 662 static inline int PageAnon(struct page *page) 663 { 664 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0; 665 } 666 667 /* 668 * Return the pagecache index of the passed page. Regular pagecache pages 669 * use ->index whereas swapcache pages use ->private 670 */ 671 static inline pgoff_t page_index(struct page *page) 672 { 673 if (unlikely(PageSwapCache(page))) 674 return page_private(page); 675 return page->index; 676 } 677 678 /* 679 * The atomic page->_mapcount, like _count, starts from -1: 680 * so that transitions both from it and to it can be tracked, 681 * using atomic_inc_and_test and atomic_add_negative(-1). 682 */ 683 static inline void reset_page_mapcount(struct page *page) 684 { 685 atomic_set(&(page)->_mapcount, -1); 686 } 687 688 static inline int page_mapcount(struct page *page) 689 { 690 return atomic_read(&(page)->_mapcount) + 1; 691 } 692 693 /* 694 * Return true if this page is mapped into pagetables. 695 */ 696 static inline int page_mapped(struct page *page) 697 { 698 return atomic_read(&(page)->_mapcount) >= 0; 699 } 700 701 /* 702 * Different kinds of faults, as returned by handle_mm_fault(). 703 * Used to decide whether a process gets delivered SIGBUS or 704 * just gets major/minor fault counters bumped up. 705 */ 706 707 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */ 708 709 #define VM_FAULT_OOM 0x0001 710 #define VM_FAULT_SIGBUS 0x0002 711 #define VM_FAULT_MAJOR 0x0004 712 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ 713 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned page */ 714 715 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ 716 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ 717 718 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON) 719 720 /* 721 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. 722 */ 723 extern void pagefault_out_of_memory(void); 724 725 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) 726 727 extern void show_free_areas(void); 728 729 int shmem_lock(struct file *file, int lock, struct user_struct *user); 730 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags); 731 int shmem_zero_setup(struct vm_area_struct *); 732 733 #ifndef CONFIG_MMU 734 extern unsigned long shmem_get_unmapped_area(struct file *file, 735 unsigned long addr, 736 unsigned long len, 737 unsigned long pgoff, 738 unsigned long flags); 739 #endif 740 741 extern int can_do_mlock(void); 742 extern int user_shm_lock(size_t, struct user_struct *); 743 extern void user_shm_unlock(size_t, struct user_struct *); 744 745 /* 746 * Parameter block passed down to zap_pte_range in exceptional cases. 747 */ 748 struct zap_details { 749 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */ 750 struct address_space *check_mapping; /* Check page->mapping if set */ 751 pgoff_t first_index; /* Lowest page->index to unmap */ 752 pgoff_t last_index; /* Highest page->index to unmap */ 753 spinlock_t *i_mmap_lock; /* For unmap_mapping_range: */ 754 unsigned long truncate_count; /* Compare vm_truncate_count */ 755 }; 756 757 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, 758 pte_t pte); 759 760 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, 761 unsigned long size); 762 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address, 763 unsigned long size, struct zap_details *); 764 unsigned long unmap_vmas(struct mmu_gather **tlb, 765 struct vm_area_struct *start_vma, unsigned long start_addr, 766 unsigned long end_addr, unsigned long *nr_accounted, 767 struct zap_details *); 768 769 /** 770 * mm_walk - callbacks for walk_page_range 771 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry 772 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry 773 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry 774 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry 775 * @pte_hole: if set, called for each hole at all levels 776 * @hugetlb_entry: if set, called for each hugetlb entry 777 * 778 * (see walk_page_range for more details) 779 */ 780 struct mm_walk { 781 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *); 782 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *); 783 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *); 784 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *); 785 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *); 786 int (*hugetlb_entry)(pte_t *, unsigned long, 787 unsigned long, unsigned long, struct mm_walk *); 788 struct mm_struct *mm; 789 void *private; 790 }; 791 792 int walk_page_range(unsigned long addr, unsigned long end, 793 struct mm_walk *walk); 794 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, 795 unsigned long end, unsigned long floor, unsigned long ceiling); 796 int copy_page_range(struct mm_struct *dst, struct mm_struct *src, 797 struct vm_area_struct *vma); 798 void unmap_mapping_range(struct address_space *mapping, 799 loff_t const holebegin, loff_t const holelen, int even_cows); 800 int follow_pfn(struct vm_area_struct *vma, unsigned long address, 801 unsigned long *pfn); 802 int follow_phys(struct vm_area_struct *vma, unsigned long address, 803 unsigned int flags, unsigned long *prot, resource_size_t *phys); 804 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, 805 void *buf, int len, int write); 806 807 static inline void unmap_shared_mapping_range(struct address_space *mapping, 808 loff_t const holebegin, loff_t const holelen) 809 { 810 unmap_mapping_range(mapping, holebegin, holelen, 0); 811 } 812 813 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new); 814 extern int vmtruncate(struct inode *inode, loff_t offset); 815 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end); 816 817 int truncate_inode_page(struct address_space *mapping, struct page *page); 818 int generic_error_remove_page(struct address_space *mapping, struct page *page); 819 820 int invalidate_inode_page(struct page *page); 821 822 #ifdef CONFIG_MMU 823 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, 824 unsigned long address, unsigned int flags); 825 #else 826 static inline int handle_mm_fault(struct mm_struct *mm, 827 struct vm_area_struct *vma, unsigned long address, 828 unsigned int flags) 829 { 830 /* should never happen if there's no MMU */ 831 BUG(); 832 return VM_FAULT_SIGBUS; 833 } 834 #endif 835 836 extern int make_pages_present(unsigned long addr, unsigned long end); 837 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write); 838 839 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, 840 unsigned long start, int nr_pages, int write, int force, 841 struct page **pages, struct vm_area_struct **vmas); 842 int get_user_pages_fast(unsigned long start, int nr_pages, int write, 843 struct page **pages); 844 struct page *get_dump_page(unsigned long addr); 845 846 extern int try_to_release_page(struct page * page, gfp_t gfp_mask); 847 extern void do_invalidatepage(struct page *page, unsigned long offset); 848 849 int __set_page_dirty_nobuffers(struct page *page); 850 int __set_page_dirty_no_writeback(struct page *page); 851 int redirty_page_for_writepage(struct writeback_control *wbc, 852 struct page *page); 853 void account_page_dirtied(struct page *page, struct address_space *mapping); 854 int set_page_dirty(struct page *page); 855 int set_page_dirty_lock(struct page *page); 856 int clear_page_dirty_for_io(struct page *page); 857 858 extern unsigned long move_page_tables(struct vm_area_struct *vma, 859 unsigned long old_addr, struct vm_area_struct *new_vma, 860 unsigned long new_addr, unsigned long len); 861 extern unsigned long do_mremap(unsigned long addr, 862 unsigned long old_len, unsigned long new_len, 863 unsigned long flags, unsigned long new_addr); 864 extern int mprotect_fixup(struct vm_area_struct *vma, 865 struct vm_area_struct **pprev, unsigned long start, 866 unsigned long end, unsigned long newflags); 867 868 /* 869 * doesn't attempt to fault and will return short. 870 */ 871 int __get_user_pages_fast(unsigned long start, int nr_pages, int write, 872 struct page **pages); 873 /* 874 * per-process(per-mm_struct) statistics. 875 */ 876 #if defined(SPLIT_RSS_COUNTING) 877 /* 878 * The mm counters are not protected by its page_table_lock, 879 * so must be incremented atomically. 880 */ 881 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 882 { 883 atomic_long_set(&mm->rss_stat.count[member], value); 884 } 885 886 unsigned long get_mm_counter(struct mm_struct *mm, int member); 887 888 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 889 { 890 atomic_long_add(value, &mm->rss_stat.count[member]); 891 } 892 893 static inline void inc_mm_counter(struct mm_struct *mm, int member) 894 { 895 atomic_long_inc(&mm->rss_stat.count[member]); 896 } 897 898 static inline void dec_mm_counter(struct mm_struct *mm, int member) 899 { 900 atomic_long_dec(&mm->rss_stat.count[member]); 901 } 902 903 #else /* !USE_SPLIT_PTLOCKS */ 904 /* 905 * The mm counters are protected by its page_table_lock, 906 * so can be incremented directly. 907 */ 908 static inline void set_mm_counter(struct mm_struct *mm, int member, long value) 909 { 910 mm->rss_stat.count[member] = value; 911 } 912 913 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) 914 { 915 return mm->rss_stat.count[member]; 916 } 917 918 static inline void add_mm_counter(struct mm_struct *mm, int member, long value) 919 { 920 mm->rss_stat.count[member] += value; 921 } 922 923 static inline void inc_mm_counter(struct mm_struct *mm, int member) 924 { 925 mm->rss_stat.count[member]++; 926 } 927 928 static inline void dec_mm_counter(struct mm_struct *mm, int member) 929 { 930 mm->rss_stat.count[member]--; 931 } 932 933 #endif /* !USE_SPLIT_PTLOCKS */ 934 935 static inline unsigned long get_mm_rss(struct mm_struct *mm) 936 { 937 return get_mm_counter(mm, MM_FILEPAGES) + 938 get_mm_counter(mm, MM_ANONPAGES); 939 } 940 941 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) 942 { 943 return max(mm->hiwater_rss, get_mm_rss(mm)); 944 } 945 946 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) 947 { 948 return max(mm->hiwater_vm, mm->total_vm); 949 } 950 951 static inline void update_hiwater_rss(struct mm_struct *mm) 952 { 953 unsigned long _rss = get_mm_rss(mm); 954 955 if ((mm)->hiwater_rss < _rss) 956 (mm)->hiwater_rss = _rss; 957 } 958 959 static inline void update_hiwater_vm(struct mm_struct *mm) 960 { 961 if (mm->hiwater_vm < mm->total_vm) 962 mm->hiwater_vm = mm->total_vm; 963 } 964 965 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, 966 struct mm_struct *mm) 967 { 968 unsigned long hiwater_rss = get_mm_hiwater_rss(mm); 969 970 if (*maxrss < hiwater_rss) 971 *maxrss = hiwater_rss; 972 } 973 974 #if defined(SPLIT_RSS_COUNTING) 975 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm); 976 #else 977 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm) 978 { 979 } 980 #endif 981 982 /* 983 * A callback you can register to apply pressure to ageable caches. 984 * 985 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'. It should 986 * look through the least-recently-used 'nr_to_scan' entries and 987 * attempt to free them up. It should return the number of objects 988 * which remain in the cache. If it returns -1, it means it cannot do 989 * any scanning at this time (eg. there is a risk of deadlock). 990 * 991 * The 'gfpmask' refers to the allocation we are currently trying to 992 * fulfil. 993 * 994 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is 995 * querying the cache size, so a fastpath for that case is appropriate. 996 */ 997 struct shrinker { 998 int (*shrink)(int nr_to_scan, gfp_t gfp_mask); 999 int seeks; /* seeks to recreate an obj */ 1000 1001 /* These are for internal use */ 1002 struct list_head list; 1003 long nr; /* objs pending delete */ 1004 }; 1005 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */ 1006 extern void register_shrinker(struct shrinker *); 1007 extern void unregister_shrinker(struct shrinker *); 1008 1009 int vma_wants_writenotify(struct vm_area_struct *vma); 1010 1011 extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl); 1012 1013 #ifdef __PAGETABLE_PUD_FOLDED 1014 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, 1015 unsigned long address) 1016 { 1017 return 0; 1018 } 1019 #else 1020 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); 1021 #endif 1022 1023 #ifdef __PAGETABLE_PMD_FOLDED 1024 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, 1025 unsigned long address) 1026 { 1027 return 0; 1028 } 1029 #else 1030 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); 1031 #endif 1032 1033 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); 1034 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); 1035 1036 /* 1037 * The following ifdef needed to get the 4level-fixup.h header to work. 1038 * Remove it when 4level-fixup.h has been removed. 1039 */ 1040 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) 1041 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) 1042 { 1043 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? 1044 NULL: pud_offset(pgd, address); 1045 } 1046 1047 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) 1048 { 1049 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? 1050 NULL: pmd_offset(pud, address); 1051 } 1052 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ 1053 1054 #if USE_SPLIT_PTLOCKS 1055 /* 1056 * We tuck a spinlock to guard each pagetable page into its struct page, 1057 * at page->private, with BUILD_BUG_ON to make sure that this will not 1058 * overflow into the next struct page (as it might with DEBUG_SPINLOCK). 1059 * When freeing, reset page->mapping so free_pages_check won't complain. 1060 */ 1061 #define __pte_lockptr(page) &((page)->ptl) 1062 #define pte_lock_init(_page) do { \ 1063 spin_lock_init(__pte_lockptr(_page)); \ 1064 } while (0) 1065 #define pte_lock_deinit(page) ((page)->mapping = NULL) 1066 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));}) 1067 #else /* !USE_SPLIT_PTLOCKS */ 1068 /* 1069 * We use mm->page_table_lock to guard all pagetable pages of the mm. 1070 */ 1071 #define pte_lock_init(page) do {} while (0) 1072 #define pte_lock_deinit(page) do {} while (0) 1073 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;}) 1074 #endif /* USE_SPLIT_PTLOCKS */ 1075 1076 static inline void pgtable_page_ctor(struct page *page) 1077 { 1078 pte_lock_init(page); 1079 inc_zone_page_state(page, NR_PAGETABLE); 1080 } 1081 1082 static inline void pgtable_page_dtor(struct page *page) 1083 { 1084 pte_lock_deinit(page); 1085 dec_zone_page_state(page, NR_PAGETABLE); 1086 } 1087 1088 #define pte_offset_map_lock(mm, pmd, address, ptlp) \ 1089 ({ \ 1090 spinlock_t *__ptl = pte_lockptr(mm, pmd); \ 1091 pte_t *__pte = pte_offset_map(pmd, address); \ 1092 *(ptlp) = __ptl; \ 1093 spin_lock(__ptl); \ 1094 __pte; \ 1095 }) 1096 1097 #define pte_unmap_unlock(pte, ptl) do { \ 1098 spin_unlock(ptl); \ 1099 pte_unmap(pte); \ 1100 } while (0) 1101 1102 #define pte_alloc_map(mm, pmd, address) \ 1103 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1104 NULL: pte_offset_map(pmd, address)) 1105 1106 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ 1107 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \ 1108 NULL: pte_offset_map_lock(mm, pmd, address, ptlp)) 1109 1110 #define pte_alloc_kernel(pmd, address) \ 1111 ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ 1112 NULL: pte_offset_kernel(pmd, address)) 1113 1114 extern void free_area_init(unsigned long * zones_size); 1115 extern void free_area_init_node(int nid, unsigned long * zones_size, 1116 unsigned long zone_start_pfn, unsigned long *zholes_size); 1117 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP 1118 /* 1119 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its 1120 * zones, allocate the backing mem_map and account for memory holes in a more 1121 * architecture independent manner. This is a substitute for creating the 1122 * zone_sizes[] and zholes_size[] arrays and passing them to 1123 * free_area_init_node() 1124 * 1125 * An architecture is expected to register range of page frames backed by 1126 * physical memory with add_active_range() before calling 1127 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic 1128 * usage, an architecture is expected to do something like 1129 * 1130 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, 1131 * max_highmem_pfn}; 1132 * for_each_valid_physical_page_range() 1133 * add_active_range(node_id, start_pfn, end_pfn) 1134 * free_area_init_nodes(max_zone_pfns); 1135 * 1136 * If the architecture guarantees that there are no holes in the ranges 1137 * registered with add_active_range(), free_bootmem_active_regions() 1138 * will call free_bootmem_node() for each registered physical page range. 1139 * Similarly sparse_memory_present_with_active_regions() calls 1140 * memory_present() for each range when SPARSEMEM is enabled. 1141 * 1142 * See mm/page_alloc.c for more information on each function exposed by 1143 * CONFIG_ARCH_POPULATES_NODE_MAP 1144 */ 1145 extern void free_area_init_nodes(unsigned long *max_zone_pfn); 1146 extern void add_active_range(unsigned int nid, unsigned long start_pfn, 1147 unsigned long end_pfn); 1148 extern void remove_active_range(unsigned int nid, unsigned long start_pfn, 1149 unsigned long end_pfn); 1150 extern void remove_all_active_ranges(void); 1151 void sort_node_map(void); 1152 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, 1153 unsigned long end_pfn); 1154 extern unsigned long absent_pages_in_range(unsigned long start_pfn, 1155 unsigned long end_pfn); 1156 extern void get_pfn_range_for_nid(unsigned int nid, 1157 unsigned long *start_pfn, unsigned long *end_pfn); 1158 extern unsigned long find_min_pfn_with_active_regions(void); 1159 extern void free_bootmem_with_active_regions(int nid, 1160 unsigned long max_low_pfn); 1161 int add_from_early_node_map(struct range *range, int az, 1162 int nr_range, int nid); 1163 void *__alloc_memory_core_early(int nodeid, u64 size, u64 align, 1164 u64 goal, u64 limit); 1165 typedef int (*work_fn_t)(unsigned long, unsigned long, void *); 1166 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data); 1167 extern void sparse_memory_present_with_active_regions(int nid); 1168 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ 1169 1170 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \ 1171 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) 1172 static inline int __early_pfn_to_nid(unsigned long pfn) 1173 { 1174 return 0; 1175 } 1176 #else 1177 /* please see mm/page_alloc.c */ 1178 extern int __meminit early_pfn_to_nid(unsigned long pfn); 1179 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID 1180 /* there is a per-arch backend function. */ 1181 extern int __meminit __early_pfn_to_nid(unsigned long pfn); 1182 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */ 1183 #endif 1184 1185 extern void set_dma_reserve(unsigned long new_dma_reserve); 1186 extern void memmap_init_zone(unsigned long, int, unsigned long, 1187 unsigned long, enum memmap_context); 1188 extern void setup_per_zone_wmarks(void); 1189 extern void calculate_zone_inactive_ratio(struct zone *zone); 1190 extern void mem_init(void); 1191 extern void __init mmap_init(void); 1192 extern void show_mem(void); 1193 extern void si_meminfo(struct sysinfo * val); 1194 extern void si_meminfo_node(struct sysinfo *val, int nid); 1195 extern int after_bootmem; 1196 1197 extern void setup_per_cpu_pageset(void); 1198 1199 extern void zone_pcp_update(struct zone *zone); 1200 1201 /* nommu.c */ 1202 extern atomic_long_t mmap_pages_allocated; 1203 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); 1204 1205 /* prio_tree.c */ 1206 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old); 1207 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *); 1208 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *); 1209 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma, 1210 struct prio_tree_iter *iter); 1211 1212 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \ 1213 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \ 1214 (vma = vma_prio_tree_next(vma, iter)); ) 1215 1216 static inline void vma_nonlinear_insert(struct vm_area_struct *vma, 1217 struct list_head *list) 1218 { 1219 vma->shared.vm_set.parent = NULL; 1220 list_add_tail(&vma->shared.vm_set.list, list); 1221 } 1222 1223 /* mmap.c */ 1224 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); 1225 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start, 1226 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert); 1227 extern struct vm_area_struct *vma_merge(struct mm_struct *, 1228 struct vm_area_struct *prev, unsigned long addr, unsigned long end, 1229 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, 1230 struct mempolicy *); 1231 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); 1232 extern int split_vma(struct mm_struct *, 1233 struct vm_area_struct *, unsigned long addr, int new_below); 1234 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); 1235 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, 1236 struct rb_node **, struct rb_node *); 1237 extern void unlink_file_vma(struct vm_area_struct *); 1238 extern struct vm_area_struct *copy_vma(struct vm_area_struct **, 1239 unsigned long addr, unsigned long len, pgoff_t pgoff); 1240 extern void exit_mmap(struct mm_struct *); 1241 1242 extern int mm_take_all_locks(struct mm_struct *mm); 1243 extern void mm_drop_all_locks(struct mm_struct *mm); 1244 1245 #ifdef CONFIG_PROC_FS 1246 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */ 1247 extern void added_exe_file_vma(struct mm_struct *mm); 1248 extern void removed_exe_file_vma(struct mm_struct *mm); 1249 #else 1250 static inline void added_exe_file_vma(struct mm_struct *mm) 1251 {} 1252 1253 static inline void removed_exe_file_vma(struct mm_struct *mm) 1254 {} 1255 #endif /* CONFIG_PROC_FS */ 1256 1257 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages); 1258 extern int install_special_mapping(struct mm_struct *mm, 1259 unsigned long addr, unsigned long len, 1260 unsigned long flags, struct page **pages); 1261 1262 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); 1263 1264 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr, 1265 unsigned long len, unsigned long prot, 1266 unsigned long flag, unsigned long pgoff); 1267 extern unsigned long mmap_region(struct file *file, unsigned long addr, 1268 unsigned long len, unsigned long flags, 1269 unsigned int vm_flags, unsigned long pgoff); 1270 1271 static inline unsigned long do_mmap(struct file *file, unsigned long addr, 1272 unsigned long len, unsigned long prot, 1273 unsigned long flag, unsigned long offset) 1274 { 1275 unsigned long ret = -EINVAL; 1276 if ((offset + PAGE_ALIGN(len)) < offset) 1277 goto out; 1278 if (!(offset & ~PAGE_MASK)) 1279 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); 1280 out: 1281 return ret; 1282 } 1283 1284 extern int do_munmap(struct mm_struct *, unsigned long, size_t); 1285 1286 extern unsigned long do_brk(unsigned long, unsigned long); 1287 1288 /* filemap.c */ 1289 extern unsigned long page_unuse(struct page *); 1290 extern void truncate_inode_pages(struct address_space *, loff_t); 1291 extern void truncate_inode_pages_range(struct address_space *, 1292 loff_t lstart, loff_t lend); 1293 1294 /* generic vm_area_ops exported for stackable file systems */ 1295 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); 1296 1297 /* mm/page-writeback.c */ 1298 int write_one_page(struct page *page, int wait); 1299 void task_dirty_inc(struct task_struct *tsk); 1300 1301 /* readahead.c */ 1302 #define VM_MAX_READAHEAD 128 /* kbytes */ 1303 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ 1304 1305 int force_page_cache_readahead(struct address_space *mapping, struct file *filp, 1306 pgoff_t offset, unsigned long nr_to_read); 1307 1308 void page_cache_sync_readahead(struct address_space *mapping, 1309 struct file_ra_state *ra, 1310 struct file *filp, 1311 pgoff_t offset, 1312 unsigned long size); 1313 1314 void page_cache_async_readahead(struct address_space *mapping, 1315 struct file_ra_state *ra, 1316 struct file *filp, 1317 struct page *pg, 1318 pgoff_t offset, 1319 unsigned long size); 1320 1321 unsigned long max_sane_readahead(unsigned long nr); 1322 unsigned long ra_submit(struct file_ra_state *ra, 1323 struct address_space *mapping, 1324 struct file *filp); 1325 1326 /* Do stack extension */ 1327 extern int expand_stack(struct vm_area_struct *vma, unsigned long address); 1328 #ifdef CONFIG_IA64 1329 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); 1330 #endif 1331 extern int expand_stack_downwards(struct vm_area_struct *vma, 1332 unsigned long address); 1333 1334 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ 1335 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); 1336 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, 1337 struct vm_area_struct **pprev); 1338 1339 /* Look up the first VMA which intersects the interval start_addr..end_addr-1, 1340 NULL if none. Assume start_addr < end_addr. */ 1341 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) 1342 { 1343 struct vm_area_struct * vma = find_vma(mm,start_addr); 1344 1345 if (vma && end_addr <= vma->vm_start) 1346 vma = NULL; 1347 return vma; 1348 } 1349 1350 static inline unsigned long vma_pages(struct vm_area_struct *vma) 1351 { 1352 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; 1353 } 1354 1355 pgprot_t vm_get_page_prot(unsigned long vm_flags); 1356 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); 1357 int remap_pfn_range(struct vm_area_struct *, unsigned long addr, 1358 unsigned long pfn, unsigned long size, pgprot_t); 1359 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); 1360 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, 1361 unsigned long pfn); 1362 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, 1363 unsigned long pfn); 1364 1365 struct page *follow_page(struct vm_area_struct *, unsigned long address, 1366 unsigned int foll_flags); 1367 #define FOLL_WRITE 0x01 /* check pte is writable */ 1368 #define FOLL_TOUCH 0x02 /* mark page accessed */ 1369 #define FOLL_GET 0x04 /* do get_page on page */ 1370 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ 1371 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ 1372 1373 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, 1374 void *data); 1375 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, 1376 unsigned long size, pte_fn_t fn, void *data); 1377 1378 #ifdef CONFIG_PROC_FS 1379 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long); 1380 #else 1381 static inline void vm_stat_account(struct mm_struct *mm, 1382 unsigned long flags, struct file *file, long pages) 1383 { 1384 } 1385 #endif /* CONFIG_PROC_FS */ 1386 1387 #ifdef CONFIG_DEBUG_PAGEALLOC 1388 extern int debug_pagealloc_enabled; 1389 1390 extern void kernel_map_pages(struct page *page, int numpages, int enable); 1391 1392 static inline void enable_debug_pagealloc(void) 1393 { 1394 debug_pagealloc_enabled = 1; 1395 } 1396 #ifdef CONFIG_HIBERNATION 1397 extern bool kernel_page_present(struct page *page); 1398 #endif /* CONFIG_HIBERNATION */ 1399 #else 1400 static inline void 1401 kernel_map_pages(struct page *page, int numpages, int enable) {} 1402 static inline void enable_debug_pagealloc(void) 1403 { 1404 } 1405 #ifdef CONFIG_HIBERNATION 1406 static inline bool kernel_page_present(struct page *page) { return true; } 1407 #endif /* CONFIG_HIBERNATION */ 1408 #endif 1409 1410 extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk); 1411 #ifdef __HAVE_ARCH_GATE_AREA 1412 int in_gate_area_no_task(unsigned long addr); 1413 int in_gate_area(struct task_struct *task, unsigned long addr); 1414 #else 1415 int in_gate_area_no_task(unsigned long addr); 1416 #define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);}) 1417 #endif /* __HAVE_ARCH_GATE_AREA */ 1418 1419 int drop_caches_sysctl_handler(struct ctl_table *, int, 1420 void __user *, size_t *, loff_t *); 1421 unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask, 1422 unsigned long lru_pages); 1423 1424 #ifndef CONFIG_MMU 1425 #define randomize_va_space 0 1426 #else 1427 extern int randomize_va_space; 1428 #endif 1429 1430 const char * arch_vma_name(struct vm_area_struct *vma); 1431 void print_vma_addr(char *prefix, unsigned long rip); 1432 1433 void sparse_mem_maps_populate_node(struct page **map_map, 1434 unsigned long pnum_begin, 1435 unsigned long pnum_end, 1436 unsigned long map_count, 1437 int nodeid); 1438 1439 struct page *sparse_mem_map_populate(unsigned long pnum, int nid); 1440 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); 1441 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); 1442 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); 1443 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); 1444 void *vmemmap_alloc_block(unsigned long size, int node); 1445 void *vmemmap_alloc_block_buf(unsigned long size, int node); 1446 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); 1447 int vmemmap_populate_basepages(struct page *start_page, 1448 unsigned long pages, int node); 1449 int vmemmap_populate(struct page *start_page, unsigned long pages, int node); 1450 void vmemmap_populate_print_last(void); 1451 1452 extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim, 1453 size_t size); 1454 extern void refund_locked_memory(struct mm_struct *mm, size_t size); 1455 1456 enum mf_flags { 1457 MF_COUNT_INCREASED = 1 << 0, 1458 }; 1459 extern void memory_failure(unsigned long pfn, int trapno); 1460 extern int __memory_failure(unsigned long pfn, int trapno, int flags); 1461 extern int unpoison_memory(unsigned long pfn); 1462 extern int sysctl_memory_failure_early_kill; 1463 extern int sysctl_memory_failure_recovery; 1464 extern void shake_page(struct page *p, int access); 1465 extern atomic_long_t mce_bad_pages; 1466 extern int soft_offline_page(struct page *page, int flags); 1467 1468 extern void dump_page(struct page *page); 1469 1470 #endif /* __KERNEL__ */ 1471 #endif /* _LINUX_MM_H */ 1472